AI Watch, Wednesday 4 November 2026Revolut cuts its AI bill tenfold

The advocate general at the Court of Justice of the European Union has delivered an opinion in the first case tied to generative AI, Next.ink and Clubic report.

In their view, when a chatbot like Gemini ingests and reuses press publications to produce its answers, it must respect publishers' neighbouring rights: the content must have been obtained lawfully, so without crossing a paywall, and the answer must not reproduce the style of the article.

The opinion does not yet bind the Court, but it traces the line the ruling is likely to follow. For publishers and AI providers alike, it is a first framework for what an assistant may lawfully do with a protected text.

Anthropic is launching Smart Reports, a tool reserved for Claude Enterprise customers that analyses how teams use AI and what it costs the company, Clubic reports. Several observers see it as an instrument for monitoring employees, even though Anthropic's documentation sets limits on the data reported.

The tool arrives as management teams look to measure the return on an AI deployment, at the risk of confusing spending oversight with control of people.

Researchers at NorForge describe an attack that hijacks the open source platform Cairn to orchestrate autonomous AI agents, ZDNet reports. These agents correlate micro-flaws without writing a single piece of malware, and are said to have already hit more than a hundred organisations for a starting cost of about $50.

The technique lowers the entry ticket for attacks that until now were reserved for well-equipped teams.

OpenAI has reportedly raised only about $11 billion in commitments of the $30 billion targeted, and must court large fortunes to complete its round, Clubic reports. At the same time, TechCrunch notes that Anthropic's compute margins are deteriorating just when they should be improving, as the company prepares its stock market debut with faster growth.

The two most prominent labs face the same equation: the cost of compute is climbing faster than revenue. The profitability question, long deferred, is settling at the centre of investor discussions.

Google has warned that its AI tools will be off by default for student accounts in its US education offering (the K-12 programme), unless an administrator turns them on, The Verge, TechCrunch and Next.ink report. One point stings: a school that refuses AI also loses access to standard Google search, now welded to answers generated by Gemini.

It becomes impossible to enable an AI-free search for these accounts. The trade-off turns an option into a requirement and leaves little room for schools keen to limit minors' exposure to these tools.

The Australian graphic design software maker Canva has announced job cuts and role changes affecting more than a fifth of its workforce, Bloomberg reports. The company presents the reshuffle as a structural adjustment forced by artificial intelligence.

The case is one of the first large layoff plans where management openly attributes the reorganisation to AI, at one of Australia's most prominent technology companies. The signal weighs on every creative profession, where assisted production is redefining the volume of work handed to teams.

AMD has delivered a sales forecast above analyst estimates, lifted by demand for data-centre processors, and is positioning itself as Nvidia's main rival, Bloomberg reports. Server maker Super Micro also raised its annual outlook, a sign of sustained demand for machines packed with AI chips.

These results contrast with the day's warnings, including one from AllianceBernstein urging investors to diversify beyond the large AI stocks. The swing between equipment makers' euphoria and fund managers' doubt sums up the market's mood.

Chinese giant Tencent launched Hy4 on Wednesday 4 November, its new flagship open-weight language model, designed to power its enterprise software ecosystem, the South China Morning Post reports. For the first time in this family, the model understands the visual: it processes images, analyses complex charts and carries out mathematical reasoning from diagrams.

Tencent highlights stronger agentic capabilities, that is, the ability to chain actions together to accomplish a task. The release aligns it with the open race already run by DeepSeek, Z.ai and Alibaba.

At the World Internet Conference in Wuzhen, Zhejiang, Alibaba chairman Joe Tsai defended open source as "the only viable path" for building a sustainable AI industry, the South China Morning Post reports. He pointed to a growing convergence of Chinese technology companies around open ecosystems, as demand for AI climbs and US pressure on access to components intensifies.

The remarks make official a strategy the group has been pushing for months. It contrasts with the lockdown practised by the large US labs.